Rl

MU

a

ARRAA

KLLUcMHR

1

iet((imat

tfm

h0

Journal of Chromatography A, 1450 (2016) 76–85

Contents lists available at ScienceDirect

Journal of Chromatography A

jo ur nal ho me pag e: www.elsev ier .com/ locate /chroma

etention behavior of lipids in reversed-phase ultrahigh-performanceiquid chromatography–electrospray ionization mass spectrometry

agdaléna Ovcacíková, Miroslav Lísa, Eva Cífková, Michal Holcapek ∗

niversity of Pardubice, Faculty of Chemical Technology, Department of Analytical Chemistry, Studentská 573, 53210 Pardubice, Czech Republic

r t i c l e i n f o

rticle history:eceived 4 March 2016eceived in revised form 10 April 2016ccepted 29 April 2016vailable online 3 May 2016

eywords:ipidsipidomicsltrahigh-performance liquidhromatographyass spectrometryuman plasmaetention behavior

a b s t r a c t

Reversed-phase ultrahigh-performance liquid chromatography (RP-UHPLC) method using two 15 cmsub–2 �m particles octadecylsilica gel columns is developed with the goal to separate and unambigu-ously identify a large number of lipid species in biological samples. The identification is performed by thecoupling with high-resolution tandem mass spectrometry (MS/MS) using quadrupole – time-of-flight(QTOF) instrument. Electrospray ionization (ESI) full scan and tandem mass spectra are measured inboth polarity modes with the mass accuracy better than 5 ppm, which provides a high confidence oflipid identification. Over 400 lipid species covering 14 polar and nonpolar lipid classes from 5 lipid cate-gories are identified in total lipid extracts of human plasma, human urine and porcine brain. The generaldependences of relative retention times on relative carbon number or relative double bond number areconstructed and fit with the second degree polynomial regression. The regular retention patterns inhomologous lipid series provide additional identification point for UHPLC/MS lipidomic analysis, whichincreases the confidence of lipid identification. The reprocessing of previously published data by our andother groups measured in the RP mode and ultrahigh-performance supercritical fluid chromatography

on the silica column shows more generic applicability of the polynomial regression for the description ofretention behavior and the prediction of retention times. The novelty of this work is the characterizationof general trends in the retention behavior of lipids within logical series with constant fatty acyl lengthor double bond number, which may be used as an additional criterion to increase the confidence of lipididentification.

© 2016 Elsevier B.V. All rights reserved.

. Introduction

Lipids fulfill multiple essential roles within all eukaryotic cellsn living organisms [1]. Living cells contain thousands of differ-nt lipid molecules that fall into eight lipid categories accordingo LIPID MAPS classification, namely fatty acyls, glycerolipidsGL), glycerophospholipids (GP), sphingolipids (SP), sterol lipidsST), prenol lipids, saccharolipids and polyketides [1–5] contain-ng many classes and subclasses. The dysregulation of the lipid

etabolism contributes to numerous serious human diseases, suchs obesity, diabetes, cardiovascular diseases and cancer. Therefore,hey are investigated as possible biomarkers of these diseases [6–9].

Lipidomic analysis starts with the liquid – liquid lipid extrac-

ion from biological materials using organic solvents. The mostrequently used extraction procedures are based on chloroform –

ethanol – water systems according to Folch et al. [10] or Bligh and

∗ Corresponding author.E-mail address: Michal.Holcapek@upce.cz (M. Holcapek).

ttp://dx.doi.org/10.1016/j.chroma.2016.04.082021-9673/© 2016 Elsevier B.V. All rights reserved.

Dyer [11], or the extraction using methyl tert-butyl ether solventinstead of chloroform [12]. Gas chromatography–mass spectrome-try is an established approach for fatty acyl profiling [13]. Variousanalytical strategies are used in the lipidomic analysis using nontar-geted and targeted lipidomic approaches [14–18]. Another possibledivision of lipidomic approaches is according to used analyticalmethodology. Shotgun lipidomics using triple quadrupole instru-ments and characteristic precursor ion and neutral loss scans[19–21] is well established approach for the fast quantitation oflipid molecular species from extracts of biological samples with-out a chromatographic separation. The second approach is the useof liquid chromatography–mass spectrometry (LC/MS) coupling,where various chromatographic modes can be selected depend-ing on the required type of separation, such as reversed-phase (RP)LC [22–25], normal-phase (NP) LC [26,27], hydrophilic interactionliquid chromatography (HILIC) [14,15], silver-ion LC [13,28,29] and

chiral LC [30,31]. The RP separation mode coupled with MS is widelyused in a comprehensive lipidomic analysis to identify individualmolecular species in different biological samples [22,32–34], wherelipids are separated according to the length of fatty acyl chains and

romat

tmiasuhicanHltgmlaraccrdTh

oobsbapoiaop

2

2

heiewwtclpcc(a(TCteh

M. Ovcacíková et al. / J. Ch

he number and position of double bonds (DB) [22]. In the RP mode,obile phases are typically composed of mixture of water contain-

ng volatile buffers and polar organic solvents, such as methanol,cetonitrile and 2-propanol. RP mode provides intra- and interclasseparation of lipid species, especially in ultrahigh-performance liq-id chromatography (UHPLC) configuration [22], but on the otherand the quantitation is more demanding, because the lipid class

nternal standards do not coelute with analytes unlike the lipidlass separation in HILIC or NP modes. NP-LC is particularly suit-ble for the separation of nonpolar lipid classes, where individualonpolar lipid classes are separated based on their polarity [27].ILIC separation allows the lipid class separation, where individual

ipid classes are separated according to their polarity and elec-rostatic interactions [14,15]. HILIC and RP modes have relativelyood complementarity of retention mechanisms, therefore variousodes of their 2D-LC coupling have been already applied for the

ipidomic analysis [35–37]. The HILIC-like separation can be alsochieved in ultrahigh-performance supercritical fluid chromatog-aphy (UHPSFC) on silica columns, but with shorter analysis timend more efficient separation [38]. The silver-ion LC is a specialhromatographic mode based on the formation of weak reversibleomplexes of silver ions with � electrons of DB, which enables theesolution of triacylglycerols (TG) and diacylglycerols (DG) isomersiffering in the number, positions and geometry of DB [13,28,29].he most demanding separation task is a chiral resolution, whichas been applied to TG enantiomers [30,31].

The main goal of our work is the study of the retention behaviorf individual lipids in RP-UHPLC to describe general dependencesf retention times on the carbon number (CN) and the DB num-er. For this purpose, RP-UHPLC method with two C18 columns ineries is optimized and coupled to high-resolution MS/MS to unam-iguously identify the large number of lipids. The retention datare collected for lipid extracts of human plasma, human urine andorcine brain samples. Individual lipid species are identified basedn accurate m/z values of their molecular adducts and character-stic fragment ions in their MS/MS spectra measured in positive-nd negative-ion modes. Relative dependences of retention timesn the CN or the DB number are fitted with the second degreeolynomial regressions.

. Material and methods

.1. Chemicals and standards

Acetonitrile, 2-propanol, methanol, (all LC/MS gradient grade),exane (HPLC grade), chloroform (HPLC grade, stabilized by 0.5–1%thanol), ammonium acetate, sodium chloride, sodium methox-de, standards of cholest-5-en-3ß-yl octadecanoate [cholesterylster (CE) 18:1] and 3ß-hydroxy-5-cholestene [cholesterol (Chol)]ere purchased from Sigma-Aldrich (St. Louis, MO, USA). Deionizedater was prepared with a Milli-Q Reference Water Purifica-

ion System (Molsheim, France). Standards of polar lipid classesontaining C18:1(9Z) fatty acyl(s), lysophosphatidylcholine (LPC),ysophosphatidylethanolamine (LPE), phosphatidylserine (PS),hosphatidylglycerol (PG), phosphatidic acid (PA), phosphatidyl-holine (PC), phosphatidylethanolamine (PE), sphingomyelin (SM),eramide (Cer), and sphingomyelin (SM d18:1/12:0), ceramideCer d18:1/12:0), cholesteryl (d7) ester (Chol d7 16:0) and oleiccid-d9 (FA d9 18:1) were purchased from Avanti Polar LipidsAlabaster, AL, USA). Nonpolar lipid standards of TG 18:1/18:1/18:1,G 19:1/19:1/19:1 and DG 18:1/18:1 were purchased from Nu-

hek Prep (Elysian, MN, USA). The lipid nomenclature followshe shorthand notation for lipid structures published by Liebischt al. [39] and the LIPID MAPS [2] classification system. Samples ofuman plasma and urine were obtained from healthy volunteers

ogr. A 1450 (2016) 76–85 77

from the research team. Porcine brain was obtained from the localstore.

2.2. Sample preparation

Blood was collected to heparin-lithium tubes and ultracen-trifuged to obtain plasma. The total lipid extracts of human plasma,human urine and porcine brain tissue were prepared accordingto Folch procedure [10] using the chloroform – methanol – watersolvent system with minor modifications [14,15]. Human plasma(50 �L) was homogenized with 3 mL of the chloroform – methanol(2:1, v/v) mixture, while porcine brain tissue (50 mg) and humanurine (2 mL) were homogenized with 6 mL of the chloroform-methanol mixture (2:1, v/v) in the ultrasonic bath at 40 ◦C for10 min. Then, deionized water (600 �L for human plasma and1200 �L for porcine brain) was added (no additional water forhuman urine), and the mixture was centrifuged at 3000 rpm for3 min under ambient conditions. The chloroform (bottom) layercontaining lipids was collected, evaporated by a gentle stream ofnitrogen and redissolved in 1 mL of the chloroform – 2-propanol(1:1, v/v) mixture for the RP-UHPLC/ESI-MS analysis.

2.3. RP-UHPLC conditions

Experiments were performed with an Agilent 1290 Infinityseries (Agilent Technologies, Santa Clara, CA, USA). Two identicalAcquity UPLC BEH C18 columns (150 mm × 2.1 mm, 1.7 �m, Waters,Milford, MA, USA) were coupled in series and used for the sepa-ration of total lipid extracts under the following conditions. Flowrate 180 �L/min, injection volume 2 �L, column temperature 40 ◦C,mobile phase gradient 0 min – 21.5% of solvent A and 78.5% of sol-vent B, 160 min – 100% of solvent B, where solvent A was 5 mmol/Laqueous ammonium acetate and solvent B was the mixture of 99.5%of acetonitrile – 2-propanol (1:2, v/v) and 0.5% water, the concen-tration of ammonium acetate in solvent B was also 5 mmol/L. Thesystem backpressure reached 1000 bar during the gradient analy-sis.

2.4. ESI-MS conditions

The hybrid QTOF mass spectrometer (micrOTOF-Q, Bruker Dal-tonics, Bremen, Germany) with an ESI source was used as thedetector under the following conditions: capillary voltage 4.5 kV,nebulizing gas pressure 1.0 bar, drying gas flow rate 8 L/min anddrying gas temperature 200 ◦C. ESI mass spectra were measuredin the range of m/z 50–1500 in positive- and negative-ion modes.Argon as the collision gas at the collision energy of 20–25 eVwas used for MS/MS experiments. MS/MS spectra are recorded inboth polarity modes using the data independent mode for all ionsexceeding the instrumental intensity threshold of 104. The externalcalibration of the mass scale was performed with sodium formateclusters before individual measurements together with the inter-nal recalibration using the most abundant known lipids. The datawere acquired using the DataAnalysis software (Bruker Daltonics).

3. Results and discussion

3.1. RP-UHPLC separation of lipids

The goal of our RP-UHPLC analysis is the identification of thelarge number of lipid species, which is then used for the studyof the retention behavior of individual lipids in logical series

with the constant number of carbon atoms or DB. For this pur-pose, we have selected the coupling of two 15 cm C18 columnswith sub–2 �m particles (150 mm × 2.1 mm, 1.7 �m) and aqueousammonium acetate – acetonitrile – 2-propanol gradient, which

78 M. Ovcacíková et al. / J. Chromat

Table 1Numbers of identified lipids in studied biological samples.

Lipid class Human plasma Human urine Porcine brain Total

FA 31 20 39 39LPC 10 0 8 11LPE 5 0 6 6SM 33 1 21 33PI 13 0 6 13PG 1 0 5 6PE 30 18 28 33PC 50 2 40 57Sulfatides 0 0 7 7Cer 0 0 13 13HexCer 0 0 12 12DG 12 0 11 15Chol + SE 22 0 9 22

pudpirhsta

tcam[etfEeaSohl

3

eouaaISbhgau3alfdt

TG 139 6 106 149Total 346 47 311 416

rovided the best performance in our previous work on 2D-LC/MSsing single C18 column in the first dimension [35]. Optimized con-itions with two C18 columns (details in Section 2) result in highereak capacity (PC = 377 calculated as Pc = 1 + (tg/1.7 × w1/2) [40])

n comparison to one C18 column (PC = 140) [40]. The number ofeal identifications in case of optimized RP separation (416) is 10%igher than the theoretical peak capacity due to the use of recon-tructed ion currents. The analysis time of 160 min is acceptable forhe study of retention behavior but not for the routine lipidomicnalysis, where faster gradients on shorter columns are preferred.

Individual lipid species are separated in the RP mode accordingo the CN and the DB number, which is defined as the equivalentarbon number (ECN) and calculated as the total CN of all fattycyls minus two times the DB number (ECN = CN-2DB). The ECNodel has been initially developed for TG and other nonpolar lipids

41–43], but it is also applicable for polar lipid classes [35,44]. Fewxceptions from this rule can be observed for phospholipids con-aining the combination of highly polyunsaturated and saturatedatty acyls, which are retained more strongly and elute in higherCN groups, e.g., PC 18:0/20:5 and PC 18:0/22:6 (both with ECN 28)lute in the group with ECN 30. The similar behavior is observed forll polar lipid classes in RP-UHPLC. Polar lipid classes (FA, LPC, LPE,M, PI, PG, PE, PC, sulfatides, Cer, and HexCer) and DG are elutedver a broad range of retention times. Nonpolar lipids (CE and TG)ave the highest retention in the RP mode, while the most polar

ysophospholipids are retained the least (Fig. 1 and Table S1).

.2. Identification of individual lipids using RP-UHPLC/ESI-MS

The identification of individual lipids is based on our previousxperiences with retention, ionization and fragmentation behaviorf various classes of lipids [6,14,15,35,38,44,45]. Table S1 lists onlynambiguously identified lipids based on retention behavior char-cteristics, accurate m/z values (better than 5 ppm in most cases)nd characteristic fragmentation behavior in both polarity modes.n total, 416 lipid species from 14 lipid classes including FA, LPC, LPE,M, PI, PG, PE, PC, sulfatides, Cer, HexCer, DG, Chol, CE and TG haveeen positively identified in total lipid extracts of human plasma,uman urine and porcine brain (Tables 1 and S1). The Venn’s dia-ram shows differences among studied samples, how many lipidsre shared among individual matrices and what identifications arenique for some samples (Fig. S2). The lipid species level (e.g., PE6:4) is the first annotation style based just on retention times andccurate m/z values in full-scan mass spectra. The fatty acyl/alkyl

evel (e.g., PE 16:0 20:4) is used when the information of attachedatty acyls is known from MS or MS/MS spectra, but without sn-ifferentiation. The slash separator (e.g., PE 16:0/20:4) indicateshe known preference of sn-position according to the shorthand

ogr. A 1450 (2016) 76–85

lipid notation recommended by Liebisch et al. [39]. Reported lipidsare also correlated with our previous papers on the lipidomic char-acterization of biological tissues [6,44,45] and body fluids [38] toconfirm the identification and avoid any error in the list of identifiedlipids.

The identification of FA, PI, PE, sulfatides and PG classes is basedon deprotonated molecules [M−H]− in negative-ion ESI mass spec-tra. The characteristic fragment ion for lipid classes containingcholine moiety (LPC, SM and PC) is m/z 184 in the positive-ionmode. Further observed ions for lipid classes containing cholinemoiety are protonated [M+H]+ and sodiated [M+Na]+ molecules inthe positive-ion mode. The neutral loss of �m/z 141 is typical forpositive-ion ESI mass spectra of PE, while negative-ion ESI-MS/MSprovides information on the position of attached fatty acyls basedon the ratio of [R1COO]−/[R2COO]− ions, where [R2COO]− ion ismore abundant. This ratio is changed for the combination of sat-urated and highly polyunsaturated (C20:5 or C22:6) fatty acyls inPE due to the formation of [Ri]− ions from [RiCOO]− ions causedby the neutral loss of carbon dioxide [46]. This approach is usedfor other phospholipid classes as well, as discussed in more detailsin our previous works [14,15]. The regioisomeric determination oflysophospholipids (LPL) is based on the knowledge of retentionorder of 1-LPL and 2-LPL standards. Observed ions for DG are pro-tonated [M+H]+ and sodiated [M+Na]+ molecules and loss of water[M+H−H2O]+ in the positive-ion mode. The characteristic fragmention for lipid species containing cholesterol (Chol and CE) is m/z369. ESI mass spectra of Cer and HexCer exhibit characteristic ionsrelated to the type of base, e.g., m/z 264 for 18:1 and m/z 266 for 18:0,which enables the accurate identification of fatty acyl position. Theinterpretation of positive-ion ESI mass spectra of TG is based on[M+NH4]+ and [M+H−RiCOOH]+ ions. In some cases, more lipidswith identical formula are reported with different retention times,such as FA 18:1 (9.8 and 10.4 min), FA 17:0 (10.6 and 11.3 min),FA 21:0 (21.2 and 22.8 min), PC 18:1/22:6, (40.0 and 40.6 min),PC 16:0/20:4 (41.0 and 44.0 min), etc. The most likely explanationfor observed isomers is different positions of DB for unsaturatedfatty acyls and branching for saturated fatty acyls, but we cannotreport this information in tables due to missing verification withstandards.

Fig. S1 shows TIC chromatograms of RP-UHPLC/ESI-MS anal-ysis of the lipid standard mixture in positive-ion (Fig. S1A) andnegative-ion (Fig. S1B) modes. The lipid standard mixture containsCE, Cer, Chol, DG, FA, GlcCer, LacCer, LPC, LPE, PA, PC, PE, PG, PS,SM and TG lipid class representatives with 18:1 fatty acyls and it isused throughout this work for the method development. Table S2reports the major ions for individual lipid classes observed in bothpolarity modes. Fig. 1 depicts TIC chromatograms of measured lipidextracts: A/human plasma containing 11 lipid classes (FA, LPC, LPE,SM, PI, PG, PE, PC, DG, CE and TG), B/human urine containing 5 lipidclasses (FA, SM, PE, PC and TG), and C/porcine brain containing 14lipid classes (FA, LPC, LPE, SM, PI, PG, PE, PC, sulfatides, Cer, HexCer,DG, CE and TG). The porcine brain extract is selected due to knownlipidomic complexity, which yields the detection of additional lipidclasses not occurring in other studied samples, i.e., sulfatides, Cerand HexCer. The type of glycosylation cannot be determined, there-fore HexCer annotation is used instead of GluCer for the standard.These RP conditions with very long gradient time are purposelyselected for the description of retention behavior of various lipidclasses, but our conclusions are also applicable for shorter gradientsused in the routine lipid analysis. This identification list will be usedas the supporting information in our future quantitative lipidomicstudies using either faster UHPLC/MS or shotgun MS approaches.

M. Ovcacíková et al. / J. Chromatogr. A 1450 (2016) 76–85 79

Fig. 1. Positive-ion RP-UHPLC/ESI-MS total ion current chromatograms of total lipid extracts of: (A) human plasma, (B) human urine, and (C) porcine brain. Conditions: twoAcquity UPLC BEH C18 columns (150 mm × 2.1 mm, 1.7 �m) coupled in series, flow rate 180 �L/min, injection volume 2 �L, column temperature 40 ◦C, mobile phase gradientof acetonitrile, 2-propanol and 5 mmol/L aqueous ammonium acetate (other details in Material and methods). FA are detected only in the negative-ion mode. Abbreviations:CE – cholesteryl ester, Cer – ceramide, DG – diacylglycerol, FA – fatty acid, HexCer – hexosyl ceramide, LPC – lysophosphatidylcholine, LPE – lysophosphatidylethanolamine,PC – phosphatidylcholine, PE – phosphatidylethanolamine, PI – phosphatidylinositol, PG – phosphatidylglycerol, SM – sphingomyelin, and TG – triacylglycerol.

80 M. Ovcacíková et al. / J. Chromatogr. A 1450 (2016) 76–85

F relativi d (F) T

3

tdprrrhadXado1D

ig. 2. Plots of polynomial dependences of relative retention times of lipids on the

s from 32 to 38), (D) Cer X:1 (X is from 34 to 42), (E) CE X:0 (X is from 14 to 18), an

.3. Study of retention behavior of various lipid classes

For all lipid classes studied, dependences of retention times onhe CN (X) or the DB number (Y) can be fitted with the secondegree polynomial regression y = ax2 + bx + c. These curves can belotted either as retention times vs. CN (or DB number) or as relativeetention times vs. relative CN (or relative DB number). The use ofelative units is more universal, because coefficients of polynomialegression do not depend on the column length or diameter. Theighest value in a particular plot is equal to 1.00 and other valuesre recalculated. Typical dependences are shown in Fig. 2 for depen-ences on the CN for A/FA X:1, B/SM X:2, C/PC X:2, D/Cer X:1, E/CE:0, and F/TG X:0. Table 2 shows that correlation coefficients R2

re mostly better than 0.99, especially for curves containing more

ata points. Fig. 3 and Table 3 summarize results of dependencesn DB number for: A/FA 18:Y, B/PI 36:Y, C/PE 40:Y, D/PC 34:Y, E/CE8:Y, and F/TG 56:Y. Correlation coefficients are slightly worse forB dependences, which is caused by a lower number of data points

e CN: (A) FA X:1 (X is from 16 to 28), (B) SM X:2 (X is from 32 to 42), (C) PC X:2 (XG X:0 (X is from 42 to 54).

per curve compared to fatty acyl chain dependences and also mul-tiple points caused by lipids with identical CN:DB composition butdifferent fatty acyl composition. Polynomial equations still enablesatisfactory prediction of retention times, because relative errors ofretention time calculations using polynomial equations are betterthan 5%, which confirms the applicability of polynomial regressionsas the supplementary identification criterion in addition to MS data.

Polynomial equations can be also applied for the prediction ofretention times of missing lipids in their logical series. Fig. 4 showsan illustration how this approach can be used for the identificationof unknown lipid inside the measured series. These lipid stan-dards are not occurring in measured biological samples, thereforethese standards are added and then experimental retention timesare correlated with predicted retention times with relative errors

lower than 5%, as illustrated on the example of TG 19:1/19:1/19:1(Fig. 4A). Mass accuracies of molecular adducts (Fig. 4B) and prod-uct ions (Fig. 4C) are better than 5 ppm, which results in very highconfidence of such identification. Other examples are shown in

M. Ovcacíková et al. / J. Chromatogr. A 1450 (2016) 76–85 81

Table 2Parameters of polynomial regressions y = ax2 + bx + c of relative retention times of individual lipid series on the relative CN.

Lipid series a b c R2 X

FA X:0 3.7991 −3.8555 1.0368 0.9949 12, 14, 15, 16, 17, 18, 20, 21,22, 23, 24, 25, 26FA X:1 4.5336 −5.0177 1.4768 0.9991 16, 17, 18, 20, 22, 24, 26, 28FA X:3 6.9790 −9.0233 3.0443 0.9910 18, 20, 22SM X:0 −1.2578 5.3851 −3.1273 1.0000 34, 36, 40SM X:1 0.3365 2.1286 −1.4651 0.9960 30, 32, 33, 34, 36, 38, 39, 40, 41, 42, 43SM X:2 1.2945 0.7492 −1.0706 0.9969 32, 34, 36, 38, 40, 42PI X:1 21.0477 −35.6225 15.5747 1.0000 32, 34, 36PE X:1 −6.9475 15.1790 −7.2317 1.0000 34, 36, 40PE X:2 −2.2415 7.0608 −3.8193 1.0000 34, 36, 40PE X:4 −11.2849 24.7328 −12.4479 1.0000 36, 38, 40PE X:5 3.0113 −1.6411 −0.3695 0.9999 36, 38, 40, 42PE X:6 44.5140 −82.0360 38.5210 1.0000 38,40,42PC X:1 6.2750 −8.6052 3.3241 0.9952 32, 34, 36PC X:2 3.3064 −2.6559 0.3511 0.9957 32, 33, 34, 35, 36, 38PC X:3 6.3464 −8.0278 2.6421 0.9626 34, 36, 38PC X:4 −8.3879 19.8431 −10.4552 0.9458 36, 38, 40Sulfatides X:1 1.6537 0.4987 −1.1487 1.0000 36, 38, 41, 42Sulfatides X:2 −1.5463 6.9934 −4.4471 1.0000 42, 43, 44Cer X:1 0.0127 2.5362 −1.5425 0.9991 34, 35, 36, 37, 38, 40, 42Cer X:2 −2.7252 7.9161 −4.1909 1.0000 36, 38, 42HexCer X:0 1.4186 −1.3316 0.9035 0.9620 36, 40, 42, 44CE X:0 −0.2553 0.9559 0.2980 0.9902 14, 15, 16, 17, 18CE X:1 −0.2321 0.9920 0.2383 0.9994 14, 16, 17, 18TG X:0 −1.5069 3.8261 −1.3203 0.9979 42, 44, 46, 48, 49, 50, 51, 52, 54TG X:1 −1.7123 4.2233 −1.5138 0.9921 44, 46, 48, 49, 50, 52, 53, 54, 56TG X:2 −1.5937 4.0467 −1.4598 0.9930 46, 48, 49, 50, 51, 52, 53, 54, 56, 58TG X:3 −0.9837 3.0144 −1.0457 0.9787 48, 49, 50, 51, 52, 53, 54, 56, 58TG X:4 −0.1748 1.8097 −0.6370 0.9219 50, 52, 54, 56TG X:5 1.6848 −1.6437 0.9517 0.9339 52, 54, 56

Table 3Parameters of polynomial regressions y = ax2 + bx + c of relative retention times of individual lipid series on the relative DB number.

Lipid series a b c R2 Y

FA 18:Y 0.5002 −1.2249 1.0007 0.9934 0, 1, 2, 3FA 20:Y 0.8193 −1.6202 0.9707 0.9882 0, 1, 2, 3, 4, 5FA 22:Y 0.7217 −1.5616 1.0041 0.9998 0, 1, 3, 4, 5, 6FA 24:Y 1.3396 −2.2112 1.0000 1.0000 0, 1, 6LPC 18:Y 0.4519 −1.0203 0.9452 0.9799 0, 1, 2LPE 18:Y −0.2105 −0.0351 1.0000 1.0000 0, 1, 2SM 34:Y −0.1743 −0.1677 1.0000 0.9830 0, 1, 2SM 40:Y −0.0247 −0.1926 1.0000 0.9870 0, 1, 2SM 42:Y 0.1313 −0.5940 1.1834 0.9783 1, 2, 3PI 36:Y 0.6095 −1.3926 1.3181 0.9863 1, 2, 3, 4PI 38:Y 0.4390 −1.5755 1.6932 0.9622 3, 4, 5, 6PE 34:Y 0.0870 −0.4319 1.0000 1.0000 0, 1, 2PE 36:Y 0.0886 −0.7954 1.1483 0.9725 1, 2, 3, 4, 5PE 38:Y 0.6683 −1.7895 1.7471 0.9208 3, 4, 5, 6PE 40:Y −0.3108 −0.4074 1.0724 0.9977 1, 2, 4, 5, 6, 7PC 32:Y 0.0229 −0.4313 1.0000 0.9895 0, 1, 2PC 34:Y 0.1973 −0.6858 0.9998 0.9878 0, 1, 2, 3PC 36:Y 0.1975 −0.9098 1.1679 0.9830 1, 4, 5PC 38:Y 0.0965 −0.8828 1.2945 0.9658 2, 3, 4, 5, 6DG 34:Y 0.0535 −0.3395 1.0000 0.9980 0, 1, 2DG 36:Y 0.1018 −0.5536 1.1673 0.9978 1, 2, 3CE 18:Y 0.0210 −0.1889 1.0000 0.9995 0, 1, 2, 3CE 20:Y −0.2639 0.1622 0.9977 0.9945 3, 4, 5TG 46:Y −0.0151 −0.1126 1.0000 0.9828 0, 1, 2TG 48:Y 0.0069 −0.1685 1.0002 0.9731 0, 1, 2, 3TG 50:Y −0.0221 −0.1716 0.9977 0.9744 0, 1, 2, 3, 4TG 52:Y 0.0014 −0.2243 0.9999 0.9854 0, 1, 2, 3, 4, 5

TC

trdt

TG 54:Y −0.0160 −0.2182

TG 56:Y −0.0263 −0.2148

able S3 for FA d9 18:1 (relative error 2.3%), SM d18:1/12:0 (4.9%),er d18:1/12:0 (1.9%) and CE d7 16:0 (1.8%).

The prediction of retention times with defined accuracy better

han 5% yields additional identification point together with accu-ate m/z values of (de)protonated molecules, molecular adducts andiagnostic fragment ions, which increases the confidence of iden-ification. The regularity in retention times of homologous series is

0.9965 0.9720 0, 2, 3, 4, 5, 61.0268 0.9856 1, 2, 3, 4, 5, 6, 7

known for long time, but we report here for the first time the well-defined identification criteria for lipids with possible applicabilityas additional identification point similarly as for LC/MS determi-

nation of compounds in forensic toxicology and doping analysis[47].

82 M. Ovcacíková et al. / J. Chromatogr. A 1450 (2016) 76–85

F relat4 nd (F

3

opdrXSiarcrbdoba

ig. 3. Plots of polynomial dependences of relative retention times of lipids on the0:Y (Y is from 1 to 7), (D) PC 34:Y (Y is from 0 to 3), (E) CE 18:Y (Y is from 0 to 3), a

.4. Comparison of retention behavior with other studies

The general applicability of our approach for the constructionf retention times dependences has been tested on RP-LC/MS datareviously published by a different group [48] and our UHPSFC/MSata [38] obtained by the chromatographic mode with the differentetention mechanism. Results are illustrated on examples of A/PC:2, B/TG X:1, C/PC 38:Y, and D/TG 58:Y in Fig. 5, Tables S4 and5 for RP-LC and of A/PC X:1, B/TG X:1, C/PC 36:Y, and D/TG 54:Yn Fig. 6, Tables S6 and S7. Polynomial regressions have a generalpplicability for the correlation of relative retention times vs. theelative CN or relative DB number. In few cases, a quadratic coeffi-ients are so small that these curves can be considered as the linearegression. Relative errors in case of UHPSFC data [38] are affectedy very small retention times (ca. 1 min) and resulting negligible

ifferences among individual TG (in the range of tens of millisec-nds), but relative errors of retention time calculation are alwaysetter than 5% without any exception. We suggest this approachs other identification point to be generally used in UHPLC/MS

ive DB number: (A) FA 18:Y (Y is from 0 to 3), (B) PI 36:Y (Y is from 1 to 4), (C) PE) TG 56:Y (Y is from 1 to 7).

and UHPSFC/MS to strengthen the advantage of chromatographicseparation for the confident lipidomic identification based on theregularity in retention times of lipid homologous series.

4. Conclusions

The RP-UHPLC/ESI-MS method enables the separation and iden-tification of large number of individual lipid species in humanplasma, human urine and porcine brain samples. This method isapplied for the study of retention behavior of various polar andnonpolar lipid classes, where polynomial dependences of rela-tive retention times both on relative CN and relative DB numberare observed. The regularity in the retention behavior of lipidhomologous series is systematically studied, which results in thesuggestion of the second degree polynomial regressions for the

description of retention patterns of lipid logical series. Polynomialregressions of retention times can be applied as an additional cri-terion for the identification of unknown lipids in addition to MSand MS/MS data, which increases the confidence of LC/MS identi-

M. Ovcacíková et al. / J. Chromatogr. A 1450 (2016) 76–85 83

Fig. 4. RP-UHPLC/ESI-MS records of TG 19:1/19:1/19:1 standard: (A) reconstructed positive-ion chromatogram, (B) ESI-MS spectrum, and (C) MS/MS spectrum of [M+NH4]+

at m/z 944.8634.

F the r( d (D)

fittwt

ig. 5. Plots of polynomial dependences of relative retention times of PC and TG onX is from 32 to 38), (B) TG X:1 (X is from 34 to 60), (C) PC 38:Y (Y is from 2 to 8), an

cation. Another possible application is the prediction of retention

imes of missing lipids in their logical series, which may bring addi-ional identifications of lipids at expected time windows, which isorthy mainly for trace species or lipids suffering from low ioniza-

ion efficiencies, because a good quality MS and MS/MS data may

elative CN (X) and the relative DB number (Y) from RP-HPLC data [48]: (A) PC X:2 TG 58:Y (Y is from 0 to 13).

not be available in such cases. The logical series with different fatty

acyl lengths are based on the higher number of data points and theyprovide slightly better correlation coefficients, therefore they arerecommended as the first choice for the identification.

84 M. Ovcacíková et al. / J. Chromatogr. A 1450 (2016) 76–85

F the rei (D) TG

A

sC

A

t0

R

[

[

[

[

[

[

[

[

[

ig. 6. Plots of polynomial dependences of relative retention times of PC and TG ons from 32 to 40), (B) TG X:1 (X is from 42 to 56), (C) PC 36:Y (Y is from 1 to 5), and

cknowledgments

This work was supported by the ERC CZ grant project LL1302ponsored by the Ministry of Education, Youth and Sports of thezech Republic.

ppendix A. Supplementary data

Supplementary data associated with this article can be found, inhe online version, at http://dx.doi.org/10.1016/j.chroma.2016.04.82.

eferences

[1] G. van Meer, D.R. Voelker, G.W. Feigenson, Membrane lipids: where they areand how they behave, Nat. Rev. Mol. Cell Biol. 9 (2008) 112–124.

[2] LIPID MAPS, LIPID MAPS Lipidomics Gateway, 2016 (accessed 2.01.16.) http://www.lipidmaps.org/.

[3] E. Fahy, S. Subramaniam, H.A. Brown, C.K. Glass, A.H. Merrill, R.C. Murphy,C.R.H. Raetz, D.W. Russell, Y. Seyama, W. Shaw, T. Shimizu, F. Spener, G. vanMeer, M.S. VanNieuwenhze, S.H. White, J.L. Witztum, E.A. Dennis, Acomprehensive classification system for lipids, Eur. J. Lipid Sci. Technol. 107(2005) 337–364.

[4] E. Fahy, S. Subramaniam, R.C. Murphy, M. Nishijima, C.R.H. Raetz, T. Shimizu,F. Spener, G. van Meer, M.J.O. Wakelam, E.A. Dennis, Update of the LIPID MAPScomprehensive classification system for lipids, J. Lipid Res. 50 (2009) S9–S14.

[5] O. Quehenberger, A.M. Armando, A.H. Brown, S.B. Milne, D.S. Myers, A.H.Merrill, S. Bandyopadhyay, K.N. Jones, S. Kelly, R.L. Shaner, C.M. Sullards, E.Wang, R.C. Murphy, R.M. Barkley, T.J. Leiker, C.R.H. Raetz, Z.Q. Guan, G.M.Laird, D.A. Six, D.W. Russell, J.G. McDonald, S. Subramaniam, E. Fahy, E.A.Dennis, Lipidomics reveals a remarkable diversity of lipids in human plasma,J. Lipid Res. 51 (2010) 3299–3305.

[6] E. Cífková, M. Holcapek, M. Lísa, D. Vrána, J. Gatek, B. Melichar, Determination

of lipidomic differences between human breast cancer and surroundingnormal tissues using HILIC-HPLC/ESI-MS and multivariate data analysis, Anal.Bioanal. Chem. 407 (2015) 991–1002.

[7] M. Holcapek, B. Cervená, E. Cífková, M. Lísa, V. Chagovets, J. Vostálová, M.Bancírová, J. Galuszka, M. Hill, Lipidomic analysis of plasma, erythrocytes and

[

lative CN (X) and the relative DB number (Y) from UHPSFC data [38]: (A) PC X:1 (X 54:Y (Y is from 0 to 6).

lipoprotein fractions of cardiovascular disease patients using UHPLC/MSMALDI-MS and multivariate data analysis, J. Chromatogr. B 990 (2015) 52–63.

[8] C. Hu, R. van der Heijden, M. Wang, J. van der Greef, T. Hankemeier, G. Xu,Analytical strategies in lipidomics and applications in disease biomarkerdiscovery, J. Chromatogr. B 877 (2009) 2836–2846.

[9] J. Suburu, Y.Q. Chen, Lipids and prostate cancer, Prostaglandins Other LipidMediators 98 (2012) 1–10.

10] J. Folch, M. Lees, G.H.S. Stanley, A simple method for the isolation andpurification of total lipids from animal tissues, J. Biol. Chem. 226 (1957)497–509.

11] E.G. Bligh, W.J. Dyer, A rapid method of total lipid extraction and purification,Can. J. Biochem. Physiol. 37 (1959) 911–917.

12] V. Matyash, G. Liebisch, T.V. Kurzchalia, A. Shevchenko, D. Schwudke, Lipidextraction by methyl-tert-butyl ether for high-throughput lipidomics, J. LipidRes. 49 (2008) 1137–1146.

13] M. Lísa, K. Netusilová, L. Franek, H. Dvoráková, V. Vrkoslav, M. Holcapek,Characterization of fatty acid and triacylglycerol composition in animal fatsusing silver-ion and non-aqueous reversed-phase high-performance liquidchromatography/mass spectrometry and gas chromatography/flameionization detection, J. Chromatogr. A 1218 (2011) 7499–7510.

14] E. Cífková, M. Holcapek, M. Lísa, M. Ovcacíková, A. Lycka, F. Lynen, P. Sandra,Nontargeted quantitation of lipid classes using hydrophilic interaction liquidchromatography–electrospray ionization mass spectrometry with singleinternal standard and response factor approach, Anal. Chem. 84 (2012)10064–10070.

15] E. Cífková, M. Holcapek, M. Lísa, Nontargeted characterization of porcineorgans using hydrophilic interaction liquid chromatography and off-linetwo-dimensional liquid chromatography–electrospray ionization massspectrometry, Lipids 48 (2013) 915–928.

16] G. Astarita, A.C. Kendall, E.A. Dennis, A. Nicolaou, Targeted lipidomicstrategies for oxygenated metabolites of polyunsaturated fatty acids, Biochim.Biophys. Acta 1851 (2015) 456–468.

17] M.A. Masood, C. Yuan, J.K. Acharya, T.D. Veenstra, J. Blonder, Quantitation ofceramide phosphorylethanolamines containing saturated and unsaturatedsphingoid base cores, Anal. Biochem. 400 (2010) 259–269.

18] C.S. Ejsing, E. Duchoslav, J. Sampaio, K. Simons, R. Bonner, C. Thiele, K. Ekroos,A. Shevchenko, Automated identification and quantification ofglycerophospholipid molecular species by multiple precursor ion scanning,

Anal. Chem. 78 (2006) 6202–6214.

19] X.L. Han, R.W. Gross, Shotgun lipidomics: electrospray ionization massspectrometric analysis and quantitation of cellular lipidomes directly fromcrude extracts of biological samples, Mass Spec. Rev. 24 (2005) 367–412.

romat

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

Acta 492 (2003) 69–82.

M. Ovcacíková et al. / J. Ch

20] X. Han, K. Yang, R.W. Gross, Multi-dimensional mass spectrometry-basedshotgun lipidomics and novel strategies for lipidomic analyses, Mass Spec.Rev. 31 (2012) 134–178.

21] C. Papan, S. Penkov, R. Herzog, C. Thiele, T. Kurzchalia, A. Shevchenko,Systematic screening for novel lipids by shotgun lipidomics, Anal. Chem. 86(2014) 2703–2710.

22] K. Sandra, A.D. Pereira, G. Vanhoenacker, F. David, P. Sandra, Comprehensiveblood plasma lipidomics by liquid chromatography/quadrupole time-of-flightmass spectrometry, J. Chromatogr. A 1217 (2010) 4087–4099.

23] K. Retra, O.B. Bleijerveld, R.A. van Gesteil, A.G.M. Tielens, J.J. van Hellemond,J.F. Brouwers, A simple and universal method for the separation andidentification of phospholipid molecular species, Rapid Commun. MassSpectrom. 22 (2008) 1853–1862.

24] M. Beccaria, G. Sullini, F. Cacciola, P. Donato, P. Dugo, L. Mondello, Highperformance characterization of triacylglycerols in milk and milk-relatedsamples by liquid chromatography and mass spectrometry, J. Chromatogr. A1360 (2014) 172–187.

25] P. Dugo, M. Beccaria, N. Fawzy, P. Donato, F. Cacciola, L. Mondello, Massspectrometric elucidation of triacylglycerol content of Brevoortia tyrannus(menhaden) oil using non-aqueous reversed-phase liquid chromatographyunder ultra high pressure conditions, J. Chromatogr. A 1259 (2012) 227–236.

26] E. Sokol, R. Almeida, H.K. Hannibal-Bach, D. Kotowska, J. Vogt, J. Baumgart, K.Kristiansen, R. Nitsch, J. Knudsen, C.S. Ejsing, Profiling of lipid species bynormal-phase liquid chromatography nanoelectrospray ionization, and iontrap–orbitrap mass spectrometry, Anal. Biochem. 443 (2013) 88–96.

27] M. Holcapek, E. Cífková, B. Cervená, M. Lísa, J. Vostalová, J. Galuszka,Determination of nonpolar and polar lipid classes in human plasma,erythrocytes and plasma lipoprotein fractions using ultrahigh-performanceliquid chromatography–mass spectrometry, J. Chromatogr. A 1377 (2015)85–91.

28] M. Holcapek, H. Dvoráková, M. Lísa, A.J. Girón, P. Sandra, J. Cvacka,Regioisomeric analysis of triacylglycerols using silver-ion liquidchromatography–atmospheric pressure chemical ionization massspectrometry: comparison of five different mass analyzers, J. Chromatogr. A1217 (2010) 8186–8194.

29] R.O. Adlof, Normal-phase separation effects with lipids on a silver ionhigh-performance liquid chromatography column, J. Chromatogr. A 764(1997) 337–340.

30] M. Lísa, M. Holcapek, Characterization of triacylglycerol enantiomers usingchiral HPLC/APCI-MS and synthesis of enantiomeric triacylglycerols, Anal.Chem. 85 (2013) 1852–1859.

31] T. Rezanka, K. Sigler, Separation of enantiomeric triacylglycerols bychiral-phase HPLC, Lipids 49 (2014) 1251–1260.

32] R. t’Kindt, L. Jorge, E. Dumont, P. Couturon, F. David, P. Sandra, K. Sandra,Profiling and characterizing skin ceramides using reversed-phase liquidchromatography-quadrupole time-of-flight mass spectrometry, Anal. Chem.84 (2012) 403–411.

33] X.L. Gao, Q.B. Zhang, D. Meng, G. Isaac, R. Zhao, T.L. Fillmore, R.K. Chu, J.Y.Zhou, K.Q. Tang, Z.P. Hu, R.J. Moore, R.D. Smith, M.G. Katze, T.O. Metz,

reversed-phase capillary ultra-performance liquid chromatography-massspectrometry (UPLC-MS) method for comprehensive top-down/bottom-uplipid profiling, Anal. Bioanal. Chem. 402 (2012) 2923–2933.

34] T. Yamada, T. Uchikata, S. Sakamoto, Y. Yokoi, E. Fukusaki, T. Bamba,Development of a lipid profiling system using reverse-phase liquid

[

ogr. A 1450 (2016) 76–85 85

chromatography coupled to high-resolution mass spectrometry with rapidpolarity switching and an automated lipid identification software, J.Chromatogr. A 1292 (2013) 211–218.

35] M. Holcapek, M. Ovcacíková, M. Lísa, E. Cífková, T. Hájek, Continuouscomprehensive two-dimensional liquid chromatography–electrosprayionization mass spectrometry of complex lipidomic samples, Anal. Bioanal.Chem. (2015) 5033–5043.

36] S.Y. Wang, J. Li, X.Z. Shi, L.Z. Qiao, X. Lu, G.W. Xu, A novel stop-flowtwo-dimensional liquid chromatography–mass spectrometry method forlipid analysis, J. Chromatogr. A 1321 (2013) 65–72.

37] P. Dugo, N. Fawzy, F. Cichello, F. Cacciola, P. Donato, L. Mondello, Stop-flowcomprehensive two-dimensional liquid chromatography combined withmass spectrometric detection for phospholipid analysis, J. Chromatogr. A1278 (2013) 46–53.

38] M. Lísa, M. Holcapek, High-throughput and comprehensive lipidomic analysisusing ultrahigh-performance supercritical fluid chromatography–massspectrometry, Anal. Chem. 87 (2015) 7187–7195.

39] G. Liebisch, J.A. Vizcaino, H. Köfeler, M. Trotzmüller, W.J. Griffiths, G. Schmitz,F. Spener, M.J.O. Wakelam, Shorthand notation for lipid structures derivedfrom mass spectrometry, J. Lipid Res. 54 (2013) 1523–1530.

40] U.D. Neue, Theory of peak capacity in gradient elution, J. Chromatogr. A 1079(2005) 153–161.

41] M. Holcapek, P. Jandera, J. Fischer, B. Prokes, Analytical monitoring of theproduction of biodiesel by high-performance liquid chromatography withvarious detection methods, J. Chromatogr. A 858 (1999) 13–31.

42] M. Holcapek, P. Jandera, P. Zderadicka, L. Hrubá, Characterization oftriacylglycerol and diacylglycerol composition of plant oils usinghigh-performance liquid chromatography–atmospheric pressure chemicalionization mass spectrometry, J. Chromatogr. A 1010 (2003) 195–215.

43] M. Lísa, M. Holcapek, Triacylglycerols profiling in plant oils important in foodindustry, dietetics and cosmetics using high-performance liquidchromatography–atmospheric pressure chemical ionization massspectrometry, J. Chromatogr. A 1198–1199 (2008) 115–130.

44] M. Lísa, E. Cífková, M. Holcapek, Lipidomic profiling of biological tissues usingoff-line two-dimensional high-performance liquid chromatography massspectrometry, J. Chromatogr. A 1218 (2011) 5146–5156.

45] E. Cífková, M. Holcapek, M. Lísa, D. Vrána, B. Melichar, V. Student, Lipidomicdifferentiation between human kidney tumors and surrounding normaltissues using HILIC-HPLC/ESI-MS and multivariate data analysis, J.Chromatogr. B 1000 (2015) 14–21.

46] O. Berdeaux, P. Juaneda, L. Martine, S. Cabaret, L. Bretillon, N. Acar,Identification and quantification of phosphatidylcholines containingvery-long-chain polyunsaturated fatty acid in bovine and human retina usingliquid chromatography/tandem mass spectrometry, J. Chromatogr. A 1217(2010) 7738–7748.

47] L. Rivier, Criteria for the identification of compounds by liquidchromatography–mass spectrometry and liquid chromatography-multiplemass spectrometry in forensic toxicology and doping analysis, Anal. Chim.

48] A. Fauland, H. Köfeler, M. Trotzmüller, A. Knopf, J. Hartler, A. Eberl, C. Chitraju,E. Lankmazr, F. Spener, A comprehensive method for lipid profiling by liquidchromatography-ion cyclotron resonance mass spectrometry, J. Lipid Res. 52(2011) 2315–2322.